train.py

from __future__ import print_function, division

from matplotlib import pyplot as plt
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.autograd import Variable
import torchvision
from torchvision import datasets, models, transforms
import argparse
import time
import os
import cv2
from PIL import Image as pil_image
from tqdm import tqdm
from network.classifier import *
from network.transform import mesonet_data_transforms
from network.mobilenetv3 import *
import pandas as pd
import numpy as np


def train_model(model, criterion, optimizer, scheduler, num_epochs=50):
	best_model_wts = model.state_dict()
	best_acc = 0.0
	bias_variance = {'train':[], 'val':[]}
	all_epoch = [i for i in range(1, num_epochs+1)]
	for epoch in range(num_epochs):
		print('Epoch {}/{}'.format(epoch+1, num_epochs))
		print('-' * 10)
		# Each epoch has a training and validation phase
		# and we want to draw a bias-variance trade-off schematic diagram
		for phase in ['train', 'val']:
			if phase == 'train':
				scheduler.step()
				model.train()
			else:
				model.eval()
			running_loss = 0.0
			running_corrects = 0.0

			for inputs, labels in dataloaders[phase]:
				inputs, labels = Variable(inputs.cuda()), Variable(labels.cuda())
				optimizer.zero_grad()

				outputs = model(inputs)
				_, preds = torch.max(outputs.data, 1)
				loss = criterion(outputs, labels)

				if phase == 'train':
					loss.backward()
					optimizer.step()

				running_loss += loss.data.item()
				running_corrects += torch.sum(preds == labels.data).to(torch.float32)
			epoch_loss = running_loss / datasets_sizes[phase]
			epoch_acc = running_corrects / datasets_sizes[phase]

			print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
			bias_variance[phase].append(1 - float(epoch_acc))
			if phase == 'val' and epoch_acc > best_acc:
				best_acc = epoch_acc
				best_model_wts = model.state_dict()
	print('Best val Acc: {:4f}'.format(best_acc))
	model.load_state_dict(best_model_wts)
	plt.plot(all_epoch, np.array(bias_variance['train']), c='red')
	plt.plot(all_epoch, np.array(bias_variance['val']), c='blue')
	plt.show()
	
	# key to csv column
	dataframe = pd.DataFrame({'training_error':np.array(bias_variance['train']), 'validation_error':np.array(bias_variance['val'])})
	dataframe.to_csv("train.csv",index=False,sep=',')


	return model



if __name__ == '__main__':
	data_dir = '.\\extended_database'
	image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x), mesonet_data_transforms[x]) for x in ['train', 'val']}
	dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x],
													batch_size=100,
													shuffle=True,
													num_workers=4) for x in ['train', 'val']}
	datasets_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}

	# model = MesoInception4()
	model = mobilenetv3()
	print(sum(param.numel() for param in model.parameters()))
	model = model.cuda()

	criterion = nn.CrossEntropyLoss()

	optimizer_ft = optim.SGD(model.parameters(), lr=0.001, momentum=0.9, weight_decay=0.001)

	exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=5, gamma=0.5)

	model = train_model(model=model, criterion=criterion, optimizer=optimizer_ft, scheduler=exp_lr_scheduler)
	torch.save(model, ".\\output\\mobilenetv3_drop_3_layer.pkl")